Living body detection method and device

Abstract

The invention provides a living body detection method and device.The living body detection method comprises the steps that a face video is collected, it is determined that the collected face video contains a front face, a random text is generated, and after the generated random text is randomly displayed in a random area of a preset screen, a user is prompted to read the displayed random text; acquiring a reading video of the displayed random text read by the user, determining that the reading video contains a human face and voice, and based on the reading video, acquiring a position feature vector facing a screen when the user reads the random text and a view angle vector of the user watching the screen; inputting the position feature vector and the view angle vector into a pre-trained living body consistency detection model to obtain a consistency value of the position feature vector and the view angle vector; and if the consistency value is not less than a preset consistency threshold value, determining that the living body detection is passed. The identification accuracy of the living body can be improved.

Description

technical field [0001] The present invention relates to the technical field of intelligent identification, and in particular, to a method and device for living body detection. Background technique [0002] With the development of the mobile Internet, the development of biometric authentication technologies has become more and more abundant, and the accuracy of biometric identification has become higher and higher, making the application scenarios of biometric authentication more and more extensive. Using biometrics for identity authentication can correctly identify real users and exclude others from impersonating real users. [0003] Liveness detection technology is an important part of biometric identity authentication technology. Liveness detection is performed through the analysis of different details of face and voice, including but not limited to: face structured light, texture, depth map, action instructions, and simple voice instructions for living detection. [000...

AI Technical Summary

Problems solved by technology

[0004] However, this liveness detection method is easy to be broken by prosthetic samples. For example, during the liveness detection process, according to the detection strategy corresponding to the liveness detection, by pre-constructing the synthetic video and / or voice corresponding to the detection strategy, the liveness detection can be performed The result is valid, which reduces the recognition accuracy

Images